Documente Academic
Documente Profesional
Documente Cultură
Andrew Rosenberg
1/1
Last Time
2/1
Today
3/1
Clustering
Clustering Clustering is an unsupervised Machine Learning application The task is to group similar entities into groups.
4/1
5/1
6/1
7/1
8/1
9/1
10 / 1
11 / 1
In Machine Learning, we optimize objective functions to nd the best solution. Maximum Likelihood (for Frequentists) Maximum A Posteriori (for Bayesians) Empirical Risk Minimization Loss function Minimization What makes a good cluster? How do we dene loss or likelihood in a clustering solution?
12 / 1
Cluster Evaluation
Intrinsic Evaluation
Evaluate the compactness of the clusters
Extrinsic Evaluation
Compare the results to some gold standard labeled data. (Not covered today)
13 / 1
Intrinsic Evaluation
IV =
C xC
d(x, c)
14 / 1
15 / 1
16 / 1
Clustering Approaches
17 / 1
Hierarchical Clustering
Recursive Partitioning
18 / 1
Hierarchical Clustering
Recursive Partitioning
19 / 1
Hierarchical Clustering
Recursive Partitioning
20 / 1
Hierarchical Clustering
Recursive Partitioning
21 / 1
Hierarchical Clustering
Recursive Partitioning
22 / 1
Hierarchical Clustering
Agglomerative Clustering
23 / 1
Hierarchical Clustering
Agglomerative Clustering
24 / 1
Hierarchical Clustering
Agglomerative Clustering
25 / 1
Hierarchical Clustering
Agglomerative Clustering
26 / 1
Hierarchical Clustering
Agglomerative Clustering
27 / 1
Hierarchical Clustering
Agglomerative Clustering
28 / 1
Hierarchical Clustering
Agglomerative Clustering
29 / 1
Hierarchical Clustering
Agglomerative Clustering
30 / 1
Hierarchical Clustering
Agglomerative Clustering
31 / 1
Hierarchical Clustering
Agglomerative Clustering
32 / 1
Hierarchical Clustering
Agglomerative Clustering
33 / 1
K-Means Clustering
K-Means clustering is a Partitional Clustering Algorithm. Identify dierent partitions of the space for a xed number of clusters Input: a value for K the number of clusters. Output: the K centers of clusters centroids
34 / 1
K-Means Clustering
35 / 1
K-Means Clustering
Algorithm: Given an integer K specifying the number of clusters. Initialize K cluster centroids
Select K points from the data set at random Select K points from the space at random
For each point in the data set, assign it to the cluster whose center it is closest to.
argminCi d(x, Ci )
36 / 1
When an assignment is changed, the sum of squared distances of the data point to its assigned cluster is reduced.
IV is reduced.
When a cluster centroid is moved the sum of squared distances of the data points within that cluster is reduced
IV is reduced.
37 / 1
K-Means Clustering
38 / 1
K-Means Clustering
39 / 1
K-Means Clustering
40 / 1
K-Means Clustering
41 / 1
K-Means Clustering
42 / 1
K-Means Clustering
43 / 1
K-Means Clustering
44 / 1
K-Means Clustering
45 / 1
K-Means Clustering
46 / 1
K-Means Clustering
47 / 1
K-Means Clustering
48 / 1
Soft K-Means
In K-means, we forced every data point to be the member of exactly one cluster. We can relax this constraint. p(x, Ci ) = p(x, Ci ) =
Based on minimizing entropy of cluster assignment.
d(x, ci ) j d(x, cj )
exp{d(x, ci )} j exp{d(x, cj )}
We still dene a cluster by a centroid, but we calculate the centroid as a weighted center of all the data points. ci =
x
x p(x, Ci ) x p(x, Ci )
Optimal? K-means approaches a local minimum, but this is not guaranteed to be globally optimal. Could you design an approach which is globally optimal? Consistent? Dierent starting clusters can lead to dierent cluster solutions
50 / 1
Optimal? K-means approaches a local minimum, but this is not guaranteed to be globally optimal. Could you design an approach which is globally optimal? Sure, in NP. Consistent? Dierent starting clusters can lead to dierent cluster solutions
51 / 1
Suboptimality in K-Means
52 / 1
Inconsistency in K-Means
53 / 1
Inconsistency in K-Means
54 / 1
Inconsistency in K-Means
55 / 1
Inconsistency in K-Means
56 / 1
More Clustering
K-Nearest Neighbors Gaussian Mixture Models Spectral Clustering We will return to these.
57 / 1
The Project
58 / 1
Research Paper
59 / 1
Game Playing Scrabble Craps Prisoners Dilemma Financials Stock Prediction Review Systems Amazon Netix Facebook
61 / 1
Project
62 / 1
Language
Topic Classication Sentiment Analysis Speech Recognition Speaker Identication Punctuation Restoration Semantic Segmentation Recognition of Emotion, Sarcasm, etc. SMS Text normalization Chat participant identication Twitter classication/threading
63 / 1
Games
Chess Checkers Poker (Poker Academy Pro) Blackjack
Video Classication
Motion classication Segmentation
64 / 1
Bye
Next
Hidden Markov Models Viterbi Decoding
65 / 1